[ovs-dev] [HELP] Question about userspace meter

txfh2007 txfh2007 at aliyun.com
Wed Oct 30 03:30:48 UTC 2019


Hi all:
    I have been testing OVS-DPDK Rate limit using meter table these days, but I have found the rate deviation is larger than kernel meter. My test result is as below:
   case 1. No Rate limit : Iperf between two VMs(on the same compute node) TCP bandwidth is above 1G 
   case 2. Rate limit: 100Mbps, burst 80Mbps , Iperf bandwith is around 80Mbps
   case 3.Rate limit: 100Mbps, burst 100Mbps , Iperf bandwith is also around 80Mbps

   the meter setting of case 2 and 3 is as below:
   case2 : meter=1 kbps burst stats bands= type=drop rate=100000 burst_size=80000
   case3 : meter=1 kbps burst stats bands= type=drop rate=100000 burst_size=100000

   the dp->meters and meter bands parameter is as below:
   case 2:
   (gdb) p *$5
          $6 = {flags = 13, n_bands = 1, max_delta_t = 800, used = 7071330038973, packet_count = 1150879, byte_count = 1652639486, bands = 0x12adf70}
  (gdb) p $5->bands[0]
          $7 = {up = {type = 1, prec_level = 0 '\000', rate = 100000, burst_size = 80000000}, bucket = 9712, packet_count = 125916, byte_count = 180814928}
    case 3:
          $3 = {flags = 13, n_bands = 1, max_delta_t = 1000, used = 7072663887432, packet_count = 6209035, byte_count = 8916116275, bands = 0x13c1990}
(gdb) p $2->bands[0]
          $4 = {up = {type = 1, prec_level = 0 '\000', rate = 100000, burst_size = 100000000}, bucket = 489312, packet_count = 728789, byte_count = 1046541004}

I don't know the reason of the test deviation and how to set burst size ? Can anyone give me a simple explanation of the relationship between rate and burst value ? 



More information about the dev mailing list